Stages are the backbone of every forecast. If they describe buyer commitments and carry probabilities you can defend with data, deals move on evidence and the forecast lines up with reality. If they describe seller tasks and carry round-number probabilities invented in a workshop, the pipeline fills with hope and the forecast misses by thirty percent every quarter. This guide walks through the full setup: naming stages, writing exit criteria, calibrating probability from your own closed-won data, and installing the hygiene rules that keep stage integrity alive after launch.
Time: 75 minutes
Stage count is the first and most consequential decision. Too few and you cannot see where deals get stuck. Too many and reps park opportunities in safe-looking middle stages to avoid scrutiny. For most B2B motions, five or six stages hit the sweet spot. Transactional segments with sub-thirty-day cycles can run four. Enterprise motions with procurement, legal, security review, and multi-stakeholder sign-off sometimes need seven, never more. The right question is not how many stages feel comfortable but how many distinct buyer commitments exist between first qualified conversation and signed contract. Count the real inflection points. If two stages have identical exit criteria, they are one stage with a label problem. If a stage has no exit criterion that observers outside the sales team could verify, it does not deserve to exist. Keep the count honest and the pipeline stays legible.
The single biggest mistake in pipeline design is naming stages after what the rep is doing. "Demo scheduled," "proposal sent," and "contract out for signature" describe seller activities. They decay within two quarters because reps learn to game them. A rep can send a proposal to someone who has not agreed to buy. A rep can schedule a demo for a prospect who has no budget. Those stages fill with ghosts. Buyer-commitment stages describe what the prospect has actually agreed to. "Qualified need," "evaluation agreed," "economic buyer engaged," "verbal commitment," "procurement in flight." These are harder to fake because they refer to real signals from the other side of the deal. Use buyer language. Use the words a prospect would recognize if you read the stage name back to them.
An exit criterion is a short list of observable, binary facts that must be true before a deal advances. Three or four criteria per stage is the right range. More than four and reps either lie or stop updating the record. Fewer than three and the gates filter nothing. Each criterion must be checkable by someone other than the rep. "Economic buyer identified by name and title" is checkable. "Rep feels confident in the deal" is not. "Signed mutual action plan on file" is checkable. "Good relationship with sponsor" is not. Store criteria as required fields on the opportunity, not as a wiki page nobody reads. When a rep tries to drag a deal to the next stage with blank required fields, the CRM blocks the move. That friction is the single highest-leverage change you can make to forecast accuracy because it kills optimism bias at the source. The deal cannot advance until the evidence exists.
Each stage carries a win probability. That number drives weighted pipeline math and feeds every forecast model downstream. Do not invent these percentages. Do not borrow the defaults your CRM ships with. Pull them from your own history. Count every opportunity that reached a stage in the last twelve months and track how many of those eventually closed won. The resulting ratio is your real conversion rate for that stage. A stage one deal that closes at twelve percent historically has a twelve percent probability, not fifteen or twenty-five, no matter what feels right. If you have no history, borrow benchmark percentages for your motion and segment as a starting point, label them explicitly as estimates, and recalibrate after your first full quarter of real data. Probability math is only as good as the inputs. Fictional probabilities produce fictional forecasts, and no amount of downstream AI, weighting, or regression can rescue a pipeline whose foundational numbers were made up.
Probability is a math input. Forecast category is a judgment overlay. The two work together but mean different things. Probability tells you the historical odds. Category tells leadership which deals belong in Commit, Best Case, Pipeline, or Omit this quarter. A late-stage deal with weak sponsor coverage still belongs in Best Case, not Commit, no matter what the probability says. A middle-stage deal with a verbal yes from the economic buyer and a signed mutual action plan might belong in Commit even if the raw probability is sixty-five percent. Train managers to use both signals. Default each stage to a category so reps have a starting point: early stages map to Pipeline, mid-stages to Best Case, late stages to Commit. Then let managers override on a per-deal basis with a logged reason. The combination produces forecasts that respect both the math and the human judgment that math alone cannot capture.
Stages that cannot be enforced are decoration. Install validation rules that prevent a rep from advancing a deal to the next stage until the exit criteria fields for the current stage are populated. The validation message should name the specific missing field so reps fix the data instead of guessing at what the system wants. Allow backward moves without validation; honest reassessment is a sign of health and should never carry friction. But log every backward move with a timestamp and reason so pattern analysis later surfaces whether a particular stage traps too many deals or a particular rep moves deals forward prematurely. Validation on the way up, logging on the way back. The system behaves like a one-way door with evidence requirements and a two-way door for correction, which is exactly how deals should actually move.
Stage definitions decay without maintenance. Automate the maintenance so reps cannot avoid it and managers do not have to nag. Flag any opportunity whose close date is in the past, any deal with no logged activity in fourteen days, any deal where the next-step field is older than seven days, and any deal that has sat in the same stage for more than twice the average time-in-stage for its segment. These flags should surface in a daily rep view and a weekly manager view, not an email nobody opens. The point is to make rot visible, not punitive. Reps who see their own stale deals will fix them before a manager asks. Hygiene rules also protect the forecast: a deal with a thirty-day-old next step is almost always slipping, whether or not the rep has admitted it in writing. Make the slippage visible and the forecast quality improves immediately.
A pipeline is not a static artifact. Buyers behave differently each quarter, products evolve, segments mature, and competitors move. Every quarter, pull the trailing ninety days of closed-won and closed-lost opportunities and ask three questions. Did deals actually move through stages in the sequence you designed, or did they skip, double back, or route around a stage entirely? Did the exit criteria you wrote predict the deals that eventually won? Did any stage become a graveyard where deals go to stall? If a stage shows a stage-to-stage conversion rate below ten percent, either the gate is too loose or the stage should not exist. If a stage shows a conversion rate above eighty percent, it may be redundant and worth collapsing into an adjacent stage. Treat the pipeline the way a product team treats a shipping product. Measure what the users actually do inside it. Iterate on the parts that break. Never consider the design finished. The pipeline is a living contract between what your sales motion claims and what the data actually shows.
Most B2B motions perform best with five or six stages. Transactional segments with sub-thirty-day cycles can run four. Enterprise motions with multi-stakeholder procurement sometimes need seven. More than seven multiplies maintenance overhead without improving forecast accuracy and typically signals that seller activities have been mistaken for buyer commitments.
Exit criteria are the short list of observable, binary facts that must be true before a deal can advance to the next stage. They matter because they convert stage names from aspirational labels into enforced gates. Without exit criteria, reps advance deals on optimism, forecasts drift from reality, and leadership loses confidence in the pipeline. With exit criteria enforced as required fields and validation rules, every stage move carries evidence, and the forecast becomes defensible.
Pull your own historical data. Count every opportunity that touched each stage in the last twelve months, including deals that eventually lost. Divide closed-won by total touched to get the historical conversion rate per stage. That number is your probability. Round to the nearest five percent to avoid false precision. If you lack history, borrow benchmark percentages for your motion and segment as placeholders, then recalibrate after your first full quarter of real data.
Use one pipeline per sales motion, not one per team or rep. If you sell to SMB and enterprise with meaningfully different cycles and buying committees, build two pipelines. If new business and expansion follow genuinely different buyer journeys, build two. Fragmenting further bloats administration, breaks cross-rep benchmarking, and makes it impossible to answer basic coverage and conversion questions.
Treat the first ninety days after launch as a stabilization window and resist edits unless something is clearly broken. After that, review quarterly. Change stages only when you have at least one full quarter of data showing a stage traps deals (conversion below ten percent) or passes everything through (conversion above eighty percent). Never edit stages mid-quarter; the break in historical reporting is worse than the original design flaw.
Probability is a math input based on historical conversion rates. Forecast category is a judgment overlay that tells leadership which deals belong in Commit, Best Case, Pipeline, or Omit this period. A late-stage deal with weak sponsor coverage can live in Best Case even though its probability is seventy percent. The two signals work together. Probability keeps the math honest. Category respects the human judgment that math alone cannot capture.
Allow skipping forward only with manager approval and a logged reason. A deal that leaps from early qualification to late-stage negotiation is either a genuine fast-track or a rep hiding missing diligence. The approval requirement surfaces the difference. Skipping backward should be routine and encouraged because it reflects honest reassessment and improves forecast quality. Track skip patterns in a weekly report so you can distinguish healthy exceptions from systemic gaming.
Strkr gives you the stage definitions, required fields, probability math, and validation rules you need to run this playbook without stitching tools together. Build once, enforce automatically, forecast with confidence.
No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.
